Binding affinity prediction with property-encoded shape distribution signatures.

Area of Science:

• Computational chemistry
• Structural biology
• Drug discovery

Background:

• Accurate prediction of protein-ligand binding affinity is crucial for drug discovery.
• Existing scoring functions often struggle with accuracy and require subjective feature selection.

Purpose of the Study:

• To develop and validate a novel computational method for predicting protein-ligand binding affinity.
• To assess the performance of the new method against established scoring functions.

Main Methods:

• Utilized property-encoded shape distributions (PESD) to generate molecular signatures.
• Employed support vector machine (SVM) techniques to build predictive models (PESD-SVM).
• Compared PESD-SVM performance against the SFCscore regression-based method.

Main Results:

• The PESD-SVM method demonstrated comparable results to SFCscore.
• A good correlation between true and predicted binding affinities was observed for complexes with dominant enthalpic contributions.
• The method requires no subjective feature selection, relying solely on PESD signatures.

Conclusions:

• The PESD-SVM approach provides a validated and effective method for predicting protein-ligand binding affinity.
• Further improvements may be achieved by incorporating entropy and solvent effects.
• This method shows promise for accelerating drug discovery and development pipelines.